Policy Futures in Education Volume 10 Number 3 2012 www.wwwords.co.uk/PFIE
Education and the Labour Market in Brazil ALBERTO DE OLIVEIRA Federal University of Rio de Janeiro, Brazil GILBERTO ABRANTES FILHO Federal Rural University of Rio de Janeiro, Brazil
ABSTRACT The aim of this article is to compare the schooling levels of individuals with the demands of the Brazilian labour market. The results demonstrate the high probability of compatibility between occupation and schooling levels. But high propensities for under-education were identified associated with skin colour and position in family. The results are consistent with Brazilian social inequality. In conclusion, although there is a dearth of qualified labour in specific segments, the widespread existence of compatibility between workers’ occupations and their schooling levels suggests that companies have adapted themselves to the educational deficiencies of their workers.
Modernisation of the production structure and favourable conditions on the international scene have enabled Brazil to register rapid economic growth since 2003. Economic euphoria, however, has rekindled dissatisfaction among corporate businessmen with the country’s dearth of qualified labour in spite of all the investments by the government and the private sector in expanding the national education system since the 1990s, especially in the higher education segment. Any allembracing perception of the current Brazilian labour market situation, however, must necessarily take into account the nature of the economic and social formative processes that typify developing countries because the Brazilian educational system has been moulded on a heritage marked by a high degree of social inequality. The aim of this article is to compare the schooling levels of individuals with the demands of the Brazilian labour market. The exercise is designed to contribute to the debate on the compatibility between the efforts dedicated to expanding the national education system and the real needs of the Brazilian economy. The study has been organised into the present introduction and four sections. The first section shows how the economic growth strategies and political arrangements have effectively determined the present design of the Brazilian education system; the second synthesises a theoretical reference framework that relates schooling, economic growth and the labour market to each other. Following that, there is a portrayal of the methodological procedures that guided the study, and the final section presents a discussion of the results obtained. Economics, Power and Education in Brazil Historically, Brazilian society has been indelibly marked by social inequality. Its economic policy has reflected the possibilities of inserting the country into the international economy against the background of a power struggle taking place among different factions of the ruling class. Accordingly, any understanding of the present design of the education system requires an investigation of the power arrangements and the economic growth strategies constructed during the course of the twentieth century.
302
http://dx.doi.org/10.2304/pfie.2012.10.3.302
Education and the Labour Market in Brazil In a similar way to the other ex-colonies of European countries, Brazil integrated itself into world capitalism by means of its exportation of agricultural goods, and that enabled agricultural oligarchies to gain firm control of the State. The production scheme involved in such an economy dispensed with the need for any qualified labour force, thus jeopardising any eventual expansion of the education system. At the very beginning of the twentieth century, teaching institutions in Brazil served a middle-class clientele made up primarily of civil servants, military personnel and traders. Those at the very top of the social pyramid received their education in Europe, while access to any form of education for the poor strata of the population was very rare. With the ending of the Second World War, the search for new markets and the benevolence of the industrialised nations stemming from the cold war created ideal conditions for industrialisation and the formation of the internal market in Brazil. The weakening of the agricultural oligarchies opened the way for an increase in industrial capitalists’ influence on the State. Urbanisation and industrial demand contributed to the organisation of the domestic market and the progress of social policies, especially those linked to social welfare, health and education. Although there was an impressive increase in education investments in Brazil after the Second World War, access to the system continued to be highly differentiated. Primary education was offered by the State, while courses for professional and technical qualification were offered by the owners’ and employers’ organisations. Formal secondary and higher education were mainly restricted to private schools. That meant that workers could only attain a primary level of education followed at the most by some kind of professional/technical training, while secondary and especially higher education were destined exclusively for those in the middle class and the higher income brackets. With the advent of the military takeover in the 1960s, the favourable performance of the Brazilian economy, and later, in the 1970s, wide access to international credit, it became possible for the State to intensify its investments. Although the education system benefited from that situation, the inequalities in access to education persisted. In spite of a considerable increase in public institutions offering secondary and higher education, only those strata of the population with middle-class incomes or higher actually managed to enter a university. In the 1980s the external debt crisis that affected almost all peripheral countries interrupted the Brazilian economy’s growth cycle. Price instability and conservative measures adopted to adjust the economy severely hampered the public sector’s investment capacity and, as a consequence, the country’s social policies. Economic expansion only began again in the 1990s under the influence of neo-classic economists who advocated in favour of the preponderance of the market as opposed to the intervention of the State. The opening up of the economy, privatisations and a rigid control of inflation by means of currency devaluation mechanisms created the conditions needed for the Brazilian economy to modernise. The price of that modernisation, however, was a serious increase in unemployment and an increasingly precarious labour market. In the 2000s, the resurgence of growth in the global economy (especially that of China) and the re-structuring of Brazilian industry have enabled Brazil to present a consistent pattern of economic growth. Against that background, neo-classical economists have been insisting on the need for investments in human capital, particularly in terms of the offer of places in higher education, if such growth is to be maintained. Since the mid-1990s the federal government has widely accepted and incorporated that thesis, but due to federal budget limitations the expansion of higher education has largely been undertaken by the private sector, albeit with the support of State subsidies. The education market has become the site of good business opportunities in Brazil. However, in spite of the proliferation of private higher education institutions, highly qualified professionals continue to be educated almost exclusively in public higher education institutions. The superior quality of the public universities in comparison with the private ones is notorious, and that is especially true in the field of those sciences where the potential for transforming knowledge into technology is greatest. Access to higher secondary education has greatly increased over the last 30 years, but it is still insufficient. Finally basic education is now available to the entire population, but there are serious criticisms of its quality. In short, while there can be no doubt about the real improvement that has taken place in the Brazilian education system, it has taken place in the shadow of continuing social inequalities. The question that must be posed is whether it is 303
Alberto de Oliveira & Gilberto Abrantes Filho appropriate for Brazil to import education models from developed countries and base its public education policies on them. Human Capital and Over-education Neo-classical economists claim that increased productivity is essential to sustained economic growth and, accordingly, propose that governments should stimulate investment in human and physical capital. The theory of human capital proposes that differences in salaries are related to the quality of labour, which in turn is based on the cognitive capacity of the individual. Furthermore, investments in formal education and training can lead to increased cognitive capacity, thereby contributing to increased productivity and more economic growth. Increased stocks of human capital will also have an influence on decision-making regarding the investments of corporations’ physical capital because modernising the production plants depends on having suitably qualified labour available. Bhagwati and Rodriguez (1975) foresaw that the low stock of human capital held by peripheral countries would create obstacles to their economic development and generate a vicious circle of stagnation. Well-qualified workers would tend to migrate to those areas where the level of remuneration was much higher and which generally speaking were already able to count on a certain abundance of human capital. Lucas (1990) showed that any increase in the stock of physical capital was generally accompanied by flows of qualified labour. Cross-section regression calculations made by Benhabib and Spiegel (1994) demonstrated that variables associated with human capital played an important role in explaining variations in international investments and growth rates. Those studies reinforced Lucas’s hypothesis whereby differences in the stocks of human capital were what restricted the influx of capital to poor countries and their economic growth. Neo-classical economists believe that skills demanded by the various work positions underlie and explain the differences in the schooling levels of individuals. Accordingly, any growth or modernisation of the economy should be accompanied by an increase in the contingent of appropriately qualified workers, and in a similar vein, the division of labour tends to increase the degree of specialisation of individuals and so there is an overall increase in the stock of qualified work posts (Trow, 1961). Such thinking however, fails to take into account the importance of externalities affecting investments in human capital. Acemoglu (1996) declared that individuals invest in human capital in order to achieve higher remuneration for their work. Companies invest in physical capital to guarantee or expand their participation in the market, but modernisation also depends on the availability of qualified labour (manpower). In an economy where the rate of corporate modernisation and the stocks of human capital are unknown, the economic agents tend to invest (in human and physical capital) on the basis of their individual forecasts. Such actions result in increased stocks of human capital even in areas that are ostensibly short of qualified labour because companies tend to complement their qualified labour needs by providing on-the-job training. Studying the Brazilian economy in the 1970s and 1980s, Lau (Lau et al, 1993) stated that economic growth would depend on four factors: physical capital; the labour force; technological progress; and human capital. The authors concluded that of the four, technical progress alone was responsible for 40% of the economic expansion registered for the period under study. Their work also pointed to increases in schooling levels as an important factor in economic growth and they concluded that an increase of one year’s schooling in the average figure for the labour force had been responsible for an increase of 21% in the gross national product (GNP). There is no consensus in the literature, however, on the existence of a relationship between education and economic growth. Adkins (1974) states that the behaviour of schooling-level indexes is not related to the performance of the economy because it is possible to observe increases in schooling levels taking place in periods of economic deceleration. Berg (1970) showed that raising schooling levels did not lead to any increase in productivity and that in fact it could actually have the opposite effect. Bowles and Gintis (1972) proposed that the increase observed in workers’ schooling levels was in alignment not with the corporate demands associated with technological
304
Education and the Labour Market in Brazil modernisation, but instead, with ideological needs whereby differences in education levels would be used to justify inequalities in the labour market. In spite of all the controversy surrounding the relations between education and economic growth, one fact is irrefutable: in the post-war period, schooling levels increased in both central and peripheral countries. In Brazil the educational system underwent rapid transformations in the course of the twentieth century. What needs to be examined is whether the increase in schooling levels was compatible with the real demands of the labour market, and that calls for a discussion of the concept of ‘over-education’. The phenomenon of over-education is typified by the existence of individuals whose schooling level is higher than that needed to meet the demands of the employment position they occupy. Such under-employment of labour – or, in other words, such inefficient allocation of human capital – is associated with costs for the individual and the companies, thereby jeopardising increased productivity and economic growth. In the literature, over-education is usually portrayed as an elevation of the stock of skills and abilities springing from the increased offer of places in higher education and from the nature of the skills being demanded by the labour market. An identification of the causes of over-education will help to define the theories to be used in explaining the phenomenon. The main references used here are: human capital theory; the theory of career mobility; competition theory; and assignment theory. In spite of eventual deficiencies in the reference frameworks, those methodologies seek to verify compatibility between schooling level and occupation. According to human capital theory, salaries are determined by the productivity of work. Productivity is explained as stemming from the schooling level and professional experience of the individual. Over-education appears to accompany an increase in the average schooling level of the populace at large. The resulting increase in the supply of qualified labour tends to lead to a reduction in salaries, enabling the companies to make use of workers to perform tasks that do not actually require the qualifications they have. That fall in the salaries of qualified workers in turn leads to a reduction in investments in human capital. However, in the opinion of human capital theoreticians, over-education is merely a temporary symptom of a momentary unbalance in the labour market. Corporations adjust their qualified labour demands according to the growth rate of the economy. Workers do the same thing in regard to their own investments in human capital. The theory of career mobility is a variant of the human capital theory, and it considers that schooling and professional experience are elements of human capital and that one can substitute for the other. Both are positively related to productivity and company profits. The theory suggests that among those entering the labour market, youngsters with a high level of formal education tend to accept positions that demand lower qualifications than they possess to enable them to acquire enough professional experience to subsequently seek higher-salaried positions. In the framework of the theory of career mobility, over-education is also viewed as a temporary phenomenon whereby, from the individual’s point of view, working in a position that is below the level of his or her schooling level is a transitory strategy to obtain professional experience and accordingly be able to achieve a better position in the labour market in the future. According to the competition theory as applied to jobs (the job competition model), workers compete in the labour market for those jobs that offer the best salaries. That means that job hierarchies are based on salary levels. On the other hand, corporations classify workers according to their productivity and the estimated expenditure involved in training them. Formal education and professional training (offered on the job by the companies) are considered as being mutually substitutable, so that any increase in the average schooling level of workers tends to reduce the amounts that companies need to spend on training. That means that in an analysis based on competition theory, the main influence of increased schooling is on the workers’ employability. Those individuals endowed with a greater amount of human capital will be in a more advantageous position than their less-qualified competitors, and that will give them access to employment positions offering higher rewards. People with low levels of qualification will have to be content with less advantageous positions or simply remain unemployed. According to competition theory, any increase in average schooling levels leads to a decrease in salary levels and consequently to a poorer return on investments made in education. Competition in the labour market, however, leads individuals to preserve their investments in human capital, thereby 305
Alberto de Oliveira & Gilberto Abrantes Filho maintaining or increasing even further the aggregate stock of human capital in the economy and generating economic costs and inequalities in the labour market. In the theory of assignment, the productivity of individuals is positively related to their schooling levels. However, it is possible for workers with identical schooling levels to show different levels of productivity because their personal characteristics may create comparative advantages in regard to a specific employment position. The problem of over-education comes about because of a failure to synchronise individuals’ talents (comparative advantages) with the nature of their occupations. The phenomenon of over-education can also be related to socio-economic and spatial factors, as the application of the job signalling mode, the spatial mobility theory and the differential overqualification theory demonstrates. According to the first, because the information on the labour market is inadequate, the companies make use of schooling level as a selection criterion. On the other hand, individuals invest in their human capital as a strategy to enhance their employability even when the job requirements are inferior to the educational qualifications they possess. In that light, the interest is actually not in the schooling level as such but in its role as a differentiating factor in the labour market. The spatial mobility theory proposes that labour markets in small cities are more liable to present the phenomenon of over-education because they are in areas where there is a much greater restriction on changing one’s job and on any new jobs appearing. Restrictions on mobility typical of small labour markets offer the companies an opportunity to squeeze salaries, and that in turn stimulates the appearance of over-education. In a similar way, the differential over-qualification theory associates the phenomenon of over-education with restrictions on the mobility of individuals. In the latter case, the restrictions are linked to characteristics of the individuals themselves. A typical illustration of that is the influence exercised on women by professional decisions made by their husbands (Frank, 1978), or the influence of the dependence of their children on women’s strategies for inserting themselves in the labour market, forcing them to look for jobs where the salaries and/or attributes are inferior to their real potential, and accordingly placing them in a situation of over-education. Finally, it must be borne in mind that the theories that attempt to explain over-education have been based on the experience and situation of the developed countries against the background of their rapid economic growth and the expansion of their social policies in the post-World War II period. In the United States, although the major surge in the offer of university places took place in the 1960s, the phenomenon was already being investigated in the previous decade when Harris (1949) suggested the possibility of a situation arising where the supply of labour with higher education qualifications exceeded the demand, while Folger and Nam (1964) identified such a surplus on the supply side of the labour market in the 1940s and 1950s. More recently it has been estimated that in the United Kingdom, 30% of the labour force is liable to be over-educated. That figure may be as high as 45% for the USA (Groot & Maassen van den Brink, 2000). The variation in those estimates may be attributable to various factors, among which are the calculation methodology used to obtain them, the period examined in the respective studies, or even discrepancies in salary levels and the levels of satisfaction of individuals (Linsley, 2005). In Australia the percentage of over-educated employees is thought to vary between 10% and 30%. The highest figures were obtained among young people and immigrants, especially those coming from Asian countries. Links between over-education, under-employment and time – that is to say, between the under-utilisation of skills and the time spent acquiring them – are related variables. Salaries and levels of job satisfaction of over-educated individuals tend to be lower than those observed for the labour market as a whole (Linsley, 2005). This brief bibliographical review is intended to identify the general principles that have guided studies addressing the compatibility of education and the labour market. Naturally, such investigations have not been restricted to the developed countries but have been conducted in developing countries as well. In Brazil, Lau et al (1993), Santos (2002) and Diaz and Machado (2008) have analysed the situation in the country as a whole, while Cavalcanti (2008) dedicated his studies to the state of Pernambuco. The results obtained for Brazil as a whole will be commented on right after the presentation of the methodological procedures adopted for the present study. 306
Education and the Labour Market in Brazil Measures Taken to Make Schooling and Occupations Compatible This study uses information drawn from the Pesquisa de Emprego e Desemprego (PED) (Employment and Unemployment Survey) between 2000 and 2007 conducted by the DIEESE (Departamento Intersindical de Estatística e Estudos Socioeconômicos) and the SEADE Foundadion (Fundação Sistema Estadual de Análise de Dados). The temporal homogeneity of the database is the main reason for choosing it as the source. The geographic crosscut takes in five large Brazilian metropolitan areas and the Federal District, each with a distinct socio-economic context. The temporal crosscut involves a comparison between the figures for the years 2000 and 2007. The Brazilian Occupations Classification system (CBO, 2002) is used to categorise the various levels of skills associated with the occupations. The level of aggregation adopted corresponds to the first digit of the CBO codification system and corresponds to ‘large group’. The large groups in turn are inserted into a hierarchy according to the skill levels involved. Generally speaking, the higherskilled occupations are associated with higher schooling levels. Studies of the compatibility between schooling and occupations may differ in regard to the calculations of the contingents making up the compatibility categories under-educated, compatibly educated and over-educated because of differences in the criteria used to establish those categories in their relationship with the schooling level required by each occupation. In the method known as job analysis (Hartog, 2000; Verhaest & Omey, 2006; Diaz & Machado, 2008, labour market analysts and specialists establish the schooling level considered appropriate and define the categories. Another method, worker self-assessment (Duncan & Hoffman, 1981; Sicherman, 1991), makes the classifications on the basis of information supplied by the job-holders themselves. Finally, there is the method of realised matches (Santos, 2002; Esteves, 2009), where the determinations are based on tendency and dispersion of the statistical distribution of schooling levels as a function of occupations, reflected in censuses and survey samples, with analyses of the central tendencies (mean, mode and median values) and the dispersion values (standard deviation, inter-quartile deviation). In the present work, as in that of Esteves (2009), three situations (categorical variableS) regarding compatibility between schooling level and occupation have been considered – namely: under-educated, compatibly educated and over-educated. Using the realised-match method, the data were grouped according to region of residence (6), year of survey (2), and occupation (10), and those groupings obeyed the parameters established by the Brazilian Classification of Occupations (CBO in Portuguese) using the maximum level of aggregation only – that is to say, large groups. That treatment of the data led to the definition of 120 data groups. Individuals with a schooling level lower than the average value minus one unit of the standard-deviation measurement for their group were classified as under-educated for their occupation, and those with a schooling level higher than the average plus one unit of the standard-deviation measurement for their group were classified as over-educated. In addition to the degrees of compatibility between jobs and schooling levels (undereducated, compatibly educated and over-educated), there are other personal characteristics of individuals that influence compatibility, such as gender, position in the family, colour of skin, ethnic group, the sector the occupation belongs to and the position held by the individual within the occupation, all of which are categorical variables. We have also taken into account characteristics that constitute continuous variables, such as the number of people living in the household, the size of the family, and the age of the individuals. The ‘square of the individual’s age’ was one of the variables adopted in order to detect any eventual concavity in the curve of influence that might represent possible inversions as a function of increased age. All the factors mentioned above are usually considered in studies designed to investigate and explain compatibility between schooling levels and occupations. The most appropriate model for handling the above-mentioned variables (categorical dependant variables, categorical independent variables and continuous variables) is the multinomial logistic regression model, also known as loglinear logit. The model uses a multinomial distribution because it is handling categorical dependent variables rather than the more normal distribution of multiple linear regression equivalents demanded by continuous variables. The model is based on the premise that a given observable categorical variable can have ‘M’ different values or possible solutions depending on the performance of its predictors, the 307
Alberto de Oliveira & Gilberto Abrantes Filho independent variables. Accordingly, the probability of a case ‘i’ presenting a given value (response) is defined by the equation: M P(i,m) = exp(zi,m) / ∑ exp(zi,m) m=1 where zi,m is the value of the ‘m-th’ variable ‘z’ (a continuous variable, not observable but obtained by observation of its predictors) as calculated for the specific case ‘i’ and its exponential is interpreted as representing the propensity of case ‘i’ to the response ‘m’. The non-observable variable ‘z’ is calculated using a composition of the values for ‘J’ available predictors. Each continuous independent variable generates a single predictor, but each categorical independent variable generates various dummy predictors that indicate absence (0) or presence (1) of given characteristics, usually resulting in: J Zi,m = ∑ bm,j xi,j j=0 where: xi,j is the observed value of the ‘j-th’ predictor for the case ‘i’; bm,j is the value of the ‘j-th’ predictor coefficient for the ‘m-th’ category obtained during processing the model – its exponential is interpreted as the relative propensity to response ‘m’ due to the predictor ‘j’; j = 0 gives the common intercept for various predictive statistic models; and in all cases, every xi,0 = 1. For each case ‘i’ the probability was calculated for each one of the ‘M’ categories up to 100%. In order to facilitate the evaluation of the results it was presumed that the model predicted correctly when in case ‘i’, the response obtained ex post facto (from the sample actually used) turned out to be the one representing the highest probability found among the ‘M’ values calculated. The statistics used to examine the model fit are obtained in the usual way by taking into account the improvement in response prediction (as compared to the prediction simply using the mode) and the contribution of each variable. The way the model was defined made it difficult to interpret the coefficients because although their magnitudes could be compared, there was no visible referential for them. To get over the practical difficulties, reference characteristics were selected for the various categorical variables involved making bmref,j = 0 and bm,jref = 0 and the intercept calculated for the model reflects that choice. Thus, one of the ‘M’ responses for the dependent variable will be elected as a reference, and in the case of each categorical independent variable, one of its predictor values (characteristics) will be chosen as the reference. In the case of the continuous variables, the natural reference would be zero, but in fact that is unnecessary because any one of its values (including zero) can be taken as a reference at any moment, according to the convenience of the analysis (zero does not always make sense in practice), or it may happen that the variable in question has values very far from zero. The case (individual) endowed with all the reference characteristics and all the independent variables will be referred to as the reference individual. The quotient P(i,m)/P(i,mref), which is now exactly equal to exp(zi,m), will now be referred to as the propensity of case ‘i’ to the response ‘m’ in order to simplify the terminology. That quotient calculated for the reference individual will in turn serve as the reference for the analysis of the effects on the probabilities of altering the characteristics of the individual. Thus, exp(bm,j), which we will refer to as the relative propensity to response ‘m’ due to the predictor ‘j’, can be readily interpreted for the categorical variables, such as how many times higher the propensity for a response ‘m’ due to a differentiated characteristic ‘j’ is than the propensity when j = reference for that category. In the case of continuous variables, exp(bm,j) expresses the relative propensity to a response ‘m’ produced by a unit increase in the variable ‘j’. Given the characteristics of the model, it must be underscored that the joint effects have a multiplying effect on their exp(bm,j), which could prove useful in an analysis of the effects of altering various characteristics simultaneously. Conclusion In this study, the categorical dependent variable – namely, compatibility – can present three different responses: under-educated (m = 1), compatibly educated (m = 2), or over-educated (m = 308
Education and the Labour Market in Brazil 3), so therefore M = 3. Considering the existence of J = 17 predictors, 13 of them refer to 5 categorical variables and 4 refer to continuous variables. All the categorical variables are dummies with xi,j = 0 attributed to absence and xi,j = 1 to the presence of a given characteristic in a categorical variable. Tables I, II and III summarise the frequencies of the variables being considered in the sample that was analysed. Compatibility Under-educated Compatibly-educated Over-educated Total
Frequency 4,219,290 18,874,210 4,482,418 27,575,918
% 15.3 68.4 16.3 100.0
Table I. Categorical dependent variable. Source: Employment and Unemployment Survey (PED).
Gender Male Female Total Position in the family Head Other Total Colour White Non-white Other (Orientals/ ethnic groups) Total Branch of activity Industry Third sector Other Total Occupation position Registered labour contract* No registered labour contract Other Total
Frequency
%
15,063,648 12,512,270 27,575,918
54.6 45.4 100.0
13,309,427 14,266,491 27,575,918
48,3 51.7 100,0
15,657,770 11,620,688 297,460
56.8 42.1 1.1
27,575,918
100.0
4,505,330 20,514,997 2,555,591 27,575,918
16.3 74.4 9.3 100.0
14,310,339
51.9
9,385,311
34.0
3,880,268 27,575,918
14.1 100.0
*In Brazil, only those workers who have their employment legally registered in companies enjoy their individual labour rights, such as paid holidays, 13th salary and salary minimums guaranteed by the labour legislation. The instrument to expedite that legal registration is the work card held by the individual. Those workers with no contract registered in their work cards are not protected by the legislation and usually occupy precarious informal positions in the labour market. Table II. Categorical variable predictors. Source: Employment and Unemployment Survey (PED).
309
Alberto de Oliveira & Gilberto Abrantes Filho
Total household residents Size of family Age Age squared
Minimum value 0 1 18 324
Maximum value 24 21 101 10,201
Average 3.92 3.77 36.40 1,471.25
Standard deviation 1.75 1.63 12.10 981.96
Table III. Continuous variable predictors. Source: Employment and Unemployment Survey (PED)
Table I shows that the results indicate the existence of a high degree of compatibility between occupations and schooling levels in the large Brazilian metropolitan areas according to the classification criteria that were adopted for the study. Considering the sample as a whole, the compatibility level was 68.4%. Situations of over-education were identified for 16.3% of individuals, and of under-education for 15.3%. It must be underscored that the application of other methodologies is liable to deliver results regarding schooling/occupation compatibility levels that are discordant with those of this study. Santos (2002) investigated the situation in Brazil using data from the National Household Sample Survey (PNAD [1] in Portuguese) for 1992, and made use of two other methodologies: (i) the criterion of Verdugo and Verdugo (1989); and (ii) the modal (variation) criterion model. That author obtained the following respective results with the two methods: under-education, 17.16% and 27.84%; compatibility, 63.26% and 40.22%; and over-education, 19.58% and 31.95%. Cavalcanti (2008), in his studies in the state of Pernambuco based on PNAD/2006 data, used the realisedmatches model to investigate mean values and inter-quartile deviations and obtained the following values: under-education, 15.0%; compatibility, 72.9%; and over-education, 12.1%. In spite of the noticeable differences in the figures themselves, the results of Santos (2002) and Cavalcanti (2008) point in the same direction as those obtained in our research. On the other hand, Diaz and Machado (2008), investigating the statistical set for Brazil as a whole from the demographic census for the year 2000, came up with very different results – namely: under-educated. 53.0%; over-educated, 17.3%; and only 28.8% in situations classified as compatible. They also registered a figure of 0.9% of individuals in an indeterminate situation. It could be that the lower compatibility figures obtained by Diaz and Machado (2008) can be attributed to their use of the job analysis model, which has a much more rigid definition of schooling than that employed by the realised-matches model, introducing differences that increase in response to the increase in the dispersions and the asymmetry of the schooling distributions. Thus, their methodological differences and variations make it very difficult to compare those apparently similar studies. Given the above provisos and reservations, we can now proceed to examine the results obtained by our research in greater detail. For the purpose of applying the linear regression model, the reference individual was attributed the following characteristics: • Dependent variable – Compatibility: m = 2, Compatible • Independent variable – Gender: j = 2, Female • Independent variable – Family position: j = 4, Other • Independent variable – Colour: j = 7, Other • Independent variable – Branch of activity: j = 10, Other • Independent variable – Position in the occupation: j = 13, Other To facilitate the analysis process we established the following reference values for the continuous variables: • Total number of household residents: sample average = 3.92 residents/household • Age: sample average = 36.40 years • Square of the age: square of the average age in the sample = 1,324.96 square years • Size of family: sample average = 3.77 members/family The statistics obtained by processing the model show that personal characteristics of individuals that were taken into account in this study resulted in a model with very poor predicting power; (even) when the data that originated the model are used as a test sample it correctly predicts 99.0% 310
Education and the Labour Market in Brazil of the cases of compatibility but only 5.7% of the cases of under-education and not a single case of over-education, bringing an overall percentage of correct prediction of 68.6% and a value for Nagelkerke’s pseudo-R2 of 0.109. Even so, the model is valid, seeing that it is a better predictor than the compatibility mode (where, in this case, compatibility corresponds to 68.4% of cases), albeit the difference is very slight and all the variables that were considered (and all their coefficients) only showed statistical significance up to the level of 1% – which points to the need to identify and include other possible explanatory variables in the model. The comparative results are set out in Tables IV-VI. The results demonstrate the high probability of compatibility between occupation and schooling levels in Brazil’s metropolitan areas. Based on the reference individual used in the study, the estimated incidence of compatibility was 72.1%. Probabilities for over-education and undereducation were estimated as 20.6% and 7.3%, respectively. However, those probabilities cannot be compared directly with the relative frequencies obtained directly from the sample (Tables III and IV) because of the differences between the characteristics of the reference individual used in applying the model and those for the sample as a whole. Independent variables
Compatibility Response: under-educated
j
Gender Position in the family Colour
Branch of activity Position in the occupation
Continuous variables
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17
Intercept Male Female Head Others White Non-white Other (Orientals/ ethnic groups) Industry Third sector Others Registered labor contract No registered labor contract Others Total residents in household Size of family Age Square of the age
Response: over-educated
b1,j
error b1,j
exp(b1,j)
b3,j
error b3,j
exp(b3,j)
-4.3320 0.0879 0.0000 0.1836 0.0000 0.9151 1.3325 0.0000
0.0095 0.0014 reference 0.0015 reference 0.0076 0.0076 reference
0.0131 1.0918 1.0000 1.2015 1.0000 2.4970 3.7906 1.0000
-1.3459 0.2172 0.0000 -0.2827 0.0000 -0.5141 -0.7805 0.0000
0.0072 0.0013 reference 0.0014 reference 0.0042 0.0043 reference
0.2603 1.2425 1.0000 0.7538 1.0000 0.5980 0.4582 1.0000
-0.1061 -0.4437 0.0000 -0.4933
0.0022 0.0018 reference 0.0018
0.8993 0.6417 1.0000 0.6106
0.3650 -0.1053 0.0000 0.5143
0.0024 0.0022 reference 0.0021
1.4404 0.9001 1.0000 1.6724
-0.0650
0.0017
0.9371
0.5585
0.0021
1.7480
0.0000 0.0437
reference 0.0007
1.0000 1.0447
0.0000 0.0206
reference 0.0007
1.0000 1.0208
0.0242 0.0497 0.0000
0.0008 0.0003 0.0000
1.0245 1.0510 1.0000
-0.0783 0.0195 -0.0003
0.0007 0.0003 0.0000
0.9247 1.0197 0.9997
Table IV. Multinomial logistic regression model (loglinear logit): parameter estimates. Source: Employment and Unemployment Survey (PED).
Characteristics Individuals Any individual
P(i,1) or P(i,under) 15.30%
Probabilities P(i,2) or P(i,comp*) 68.44%
P(I,3) or P(i,over) 16.26%
*Compatible (Comp) is the modal category. Table V. Probabilities based on frequency distribution alone. Source: Employment and Unemployment Survey (PED).
311
Alberto de Oliveira & Gilberto Abrantes Filho Probabilities Characteristics Individuals Individual “i = ref” = > Individual “i” = > Male Head White Non-white Industry Third sector Reg, contract No reg. contract +10 residents/family +10 years
P (i,1) or P (i, under) 7.31% 7.55% 9.10% 17.77% 23.35% 6.06% 4.92% 4.02% 5.96% 14.68% 11.01%
Relative propensity P (i,2) or P (i,3) or Under Exp Over Exp P (i, comp) P (i,over) (b1,j) (b3,j) 72.12% 20.58% … … Alternating each characteristic, one at a time 68.26% 24.20% 1.0918 1.2425 74.81% 16.09% 1.2015 0.7538 70.25% 11.99% 2.4970 0.5980 66.02% 8.63% 3.7906 0.4582 66.58% 27.36% 0.8993 1.4404 75.66% 19.43% 0.6417 0.9001 64.98% 32.00% 0.6106 1.6724 62.75% 31.29% 0.9371 1.7480 73.53% 11.79% 1.9716 0.5619 66.08% 22.91% 1.6446 1.2155
Table VI. Probabilities and propensities obtained using the model. Source: Employment and Unemployment Survey (PED).
High propensities for sub-education were identified, especially associated with the variables skin colour and position in the family. The highest propensities, in descending order, are for non-whites, whites and heads of families. The presence of those characteristics raises the propensity for subeducation and at the same time reduces the propensity for over-education. As the composition of the group ‘others’ is largely made up of individuals of oriental extraction it could be argued that they receive a more suitable education orientated towards a career and in greater proportions. At the other extreme, among non-whites, the educational deficiency is much higher. Heads of family who take on that position early in life are liable to interrupt their schooling/instruction because of it, while at the same time developing themselves by taking on more complex occupations attracted by better positions (salaries). It should be underscored that the results obtained for family position and colour of skin are consistent with the social inequality and distortions to be found in the Brazilian education system. The deficiencies in the education of non-white people reflect their high presence in the make-up of the poorest population strata. For heads of families with low schooling levels, the possibilities of ascending professionally lie in learning undertaken in the workplace, but that process is frequently interrupted by the phenomenon of the high job-turnover rate in the Brazilian labour market. As for the question of gender, men are more liable to be either over-educated or undereducated, with the former situation predominating. In both situations compatibility is reduced, and that shows that women tend to have an education more compatible with their occupations. It must be remembered that greater compatibility does not mean higher schooling levels. The greatest effects made visible by the calculations were those associated with the continuous variables, but it should be remembered that long intervals of time are what make them relevant, with the exception of the variable ‘square of the age’, where effects were very much reduced. Accordingly, it was found that: • The addition of 10 residents to the household or to the number of family members raises the propensity for under-education 1.97 times and reduces the propensity to over-education 0.56 times. That is perfectly coherent with the situation usually observed in the poorer strata of the population where families tend to be larger, and access to education and other public services restricted. • An increase of 10 years in the age variable raises the propensity for under-education 1.64 times and the liability to over-education 1.21 times. In the former case, that is presumed to reflect the continuous evolution of the individual’s career, involving ever more complex occupations without any accompanying evolution or change in schooling level, and in the latter case, it reflects the opposite situation – that is, the late acquisition of additional schooling by individuals stabilised in occupations of low complexity.
312
Education and the Labour Market in Brazil With regard to the position occupied by individuals within their occupations, both those with legally registered labour contracts and those in informal employment situations show a high propensity for over-education and a low propensity for under-education, and that is because the reference category used is a very diverse aggregation of casual workers with very low schooling and qualification levels. What is more noteworthy is that those workers with no legally registered labour contract are more liable to be under-educated and in general show less compatibility between their schooling and their occupations than those with registered contracts. Once again the fact of such discrepancies between workers with registered labour contracts and those with none is basically a reflection of the heterogeneity of the Brazilian labour market. With regard to the sector of activities, industrial workers are more liable to be undereducated for their posts, which may possibly be explained by the higher degree of technical specialisation demanded by the sector, and that is usually addressed by providing additional instruction. Workers in the third sector are much less liable to be over-educated or under-educated and therefore generally present a high degree of compatibility. In conclusion, although it is true that there is a dearth of qualified labour in certain specific segments, the widespread existence of a high degree of compatibility between workers’ occupations and their schooling levels (acquired in an education system with serious distortions) suggests that companies have adapted themselves to the educational deficiencies of their workers. Such adaptations seem to be a direct consequence of the socially excluding economic development model adopted in Brazil and consequently adopted by the education system as well. As a result, in periods of rapid economic growth like the one Brazil has been experiencing recently, the heritage of social inequality appears as an insuperable obstacle in the path of social and economic progress. In short, investments in the national education system need to be planned in the perspective of a sweeping long-term social policy, abandoning the practice of sporadic actions designed merely to address and foster the short-term interests of the market. Note [1] The National Household Sample Survey (PNAD in Portuguese) is a survey conducted by the Brazilian Geography and Statistics Institute (IBGE) that collects information on various aspects of Brazilian society, among them the labour market. http://www.ibge.gov.br/home/estatistica/populacao/trabalhoerendimento/pnad2008/default.sht m
References Acemoglu, D. (1996) A Microfoundation for Social Increasing Returns in Human Capital Accumulation, Quarterly Journal of Economics, 111(3), 779-804. http://dx.doi.org/10.2307/2946672 Adkins, D.L. (1974) The American Educated Labor Force: an empirical look at theories of its formation and composition, in M.S. Gordon (Ed.) Higher Education and the Labor Market. Berkeley, CA: Carnegie Commission on Higher Education. Benhabib, J. & Spiegel, M.M. (1994) The Role of Human Capital in Economic Development: evidence from aggregate cross-country data, Journal of Monetary Economics, 28, 143-173. http://dx.doi.org/10.1016/0304-3932(94)90047-7 Berg, I (1970) Education and Jobs: the great training robbery. New York: Fredric A. Praeger. Bhagwati, J. & Rodriguez, C.A. (1975) Welfare-theoretical Analyses of the Brain Drain, Journal of Development Economics, 2, 195-221. http://dx.doi.org/10.1016/0304-3878(75)90002-4 Bowles, S & Gintis, H. (1972) IQ in the US Class Structure, Social Policy, Nov./Dec. & Jan./Feb., 65-96. Cavalcanti, M.F.A. (2008) Overeducation e undereducation em Pernambuco: uma avaliação empírica. Dissertation submitted to the PIMES – Post-Graduate Economics Program at the Federal University of Pernambuco, Recife, Brazil. Classificação Brasileira de Ocupações (CBO) (2002) http://www.mtecbo.gov.br/cbosite/pages/home.jsf (accessed 8 February 2010). Diaz, M.D.M. & Machado, L. (2008) Overeducation e undereducation no Brasil: incidência e retornos, Estudos Econômicos, 38(3), 431-460. http://dx.doi.org/10.1590/S0101-41612008000300001
313
Alberto de Oliveira & Gilberto Abrantes Filho Duncan, G. & Hoffman, S.D. (1981) The Incidence and Wage Effects of Overeducation, Economics of Education Review, 1(1), 75-86. http://dx.doi.org/10.1016/0272-7757(81)90028-5 Esteves, L.A. (2009) Incompatibilidade Escolaridade-Ocupação e Salários: Evidências de uma Empresa, Revista Brasileira de Economia, 63(2), 77-90. Folger, John K. & Nam, Charles B. (1964) Trends in Education in Relation to the Occupational Structure, Sociology of Education, 38, 19-33. http://dx.doi.org/10.2307/2111816 Frank, R.H. (1978) Why Women Earn Less: the theory and estimation of differential overqualification, American Economic Review, 68, 360-373. Groot, W. & Maassen van den Brink, H. (2000) Overeducation in the Labor Market: a meta-analysis, Economics of Education Review, 19, 149-158. http://dx.doi.org/10.1016/S0272-7757(99)00057-6 Harris, Seymour (1949) The Market for College Graduates and Related Aspects of Education and Income. Cambridge, MA: Harvard University Press. Hartog, J. (2000) Over-education and Earnings: where are we, where should we go? Economics of Education Review, 19(2), 131-147. http://dx.doi.org/10.1016/S0272-7757(99)00050-3 Lau, Lawrence, Jamison, Dean, Liu, Shu-Cheng & Rivkin, Steven (1993) Education and Economic Growth: some cross-sectional evidence from Brazil, Journal of Development Economics, 41, 45-70. http://dx.doi.org/10.1016/0304-3878(93)90036-M Linsley, I. (2005) Causes of Overeducation in the Australian Labour Market, Australian Journal of Labour Economics, 8(2), 121-144. Lucas, Robert E. (1990) Why Doesn't Capital Flow from Rich to Poor Countries? American Economic Review Papers and Proceedings, LXXX, 92-96. Pesquisa de Emprego & Desemprego (PED) [Employment & Unemployment Survey] (2010) http://www.dieese.org.br/ped/bd/info.xml (accessed 10 February 2010). Santos, A.M. (2002) Overeducation no mercado de trabalho brasileiro, Revista Brasileira de Economia de Empresas, 2(2), 61-80. Sicherman, N. (1991) Overeducation in the Labor Market, Journal of Labor Economics, 9(2), 101-122. http://dx.doi.org/10.1086/298261 Trow, Martin (1961) The Second Transformation of American Education, International Journal of Comparative Sociology, 2, 146-166. http://dx.doi.org/10.1177/002071526100200202 Verdugo, R.R. & Verdugo, N.T. (1989) The Impact of Surplus Schooling on Earnings: some additional findings, Journal of Human Resources, 24(4), 629-643. http://dx.doi.org/10.2307/145998 Verhaest, D. & Omey, E. (2006) Discriminating between Alternative Measures of Over-education, Applied Economics, 38(18), 2113-2120. http://dx.doi.org/10.1080/00036840500427387
ALBERTO DE OLIVEIRA is an economist and professor at the Institute of Urban and Regional Planning and Research at the Federal University of Rio de Janeiro (IPPUR/UFRJ). He is interested in the relationship between urban and regional development and the labour market; and territorial policies and instruments of policy evaluation. Correspondence:
[email protected] GILBERTO ABRANTES FILHO is a professor at the Federal Rural University of Rio de Janeiro (UFRRJ). Correspondence:
[email protected]
314
